Hi,

I've recompiled the code for my library to redirect all calls to malloc(),
free(), realloc(), and calloc() to HeapAlloc(), HeapFree(), and HeapReAlloc().

The problem? It doesn't work -- it throws OutOfMemory when the program is
starting (in the mallocNoSync() function) even though those functions do not
fail (I've checked the return values of the functions, and they don't fail;
I've also told them to throw exceptions on failure, and none is thrown).

The funny thing is, if I redirect DMD's malloc() to the msvcrt.dll malloc(), it works perfectly fine, so it's not like it's a library issue. And it doesn't seem to be an alignment issue either. Does anyone have any idea what might cause this? (And I've redirected *all* calls, including the ones from monitor.c and such, so it isn't a problem with one allocator freeing another allocator's memory.)

What could possibly be the issue? Any ideas?

Thank you!

On Tue, 04 Jan 2011 01:12:53 +0200, %u <wfunction@hotmail.com> wrote:

>
> I've recompiled the code for my library to redirect all calls to malloc(),
> free(), realloc(), and calloc() to HeapAlloc(), HeapFree(), and HeapReAlloc().

Sorry, do you mean the libc functions? The GC doesn't use those to allocate its memory pools, and uses page allocation functions (mmap on *nix and VirtualAlloc on Windows). If you mean the GC functions, then wouldn't your hooked malloc be called before mallocNoSync is called?

-- 
Best regards,
 Vladimir                            mailto:vladimir@thecybershadow.net

> Sorry, do you mean the libc functions? The GC doesn't use those to allocate its memory pools, and uses page allocation functions (mmap on *nix and VirtualAlloc on Windows). If you mean the GC functions, then wouldn't your hooked malloc be called before mallocNoSync is called?

I actually found the error (after > 8 hours of searching); it was because HeapReAlloc wasn't calling HeapAlloc if the passed pointer was null, the way realloc() is supposed to. Thank you for your reply though!